Business rules: Decisions that can be explained.

As promised in a previous article, here is a use case of streaming data in real time and how we could benefit from it.
Data science is a large domain, our hands can only count a limited amount of ways to analyze data. There are sub-categories of DS that should get more attention, especially with the potential it has.

Business rules are one of these. Defining this term in a simple phrase is complicated, let us try anyways!

Input data flows through a set of user-defined conditions and constraints until it reaches the output node: The decision to make.

Everything can be explained with facts. It goes beyond than the interpretability of a linear/logistic regression. We have the tools to chain several decisions depending on the state of the system.
This is important if we want to convince stakeholders to move forward in a project.

The data scientist inside of us might think the following: “Eww! I can’t use my Random Forest to solve this problem?”. Hang on with me for a second, you will see where we are headed.

By taking the first step into business rules, a whole set of possibilities opens. Currently, a lot of companies have data that is not usable in machine learning models. They are asking themselves how they can bridge this gap.
Business rules enables us to define a framework where we can implement models from our favorite library. The model can then help us in creating new rules. It still requires human supervision, but it is better than staring at our data.

Being compatible with IoT devices, business rules are useful in a context where a lot of data is generated each second, a quick action is of the utmost importance. By using recent innovations in database systems (MongoDB, RedisDB, PostgreSQL, etc.), the programming logic can be done inside the database. Scaling does not become an issue in this case.

Explorai recommends MongoDB as a database for this type of project. You can read more about my certification here.

The steps to defines business rules was not mentioned in this article: It is entirely dependent on the problem we are working on. It is nevertheless important to keep in mind the following:

· Specify the data to extract depending on the inputs. We do not want to copy in the memory the entire database each time the state of the system changes. With IoT devices, updates can be done several times in the same second!

· Configuration of rules inside the database: The programming is shared between the application and the database.

· A decision can imply a change in the database: One of the actions could be to write a new value in the database, thus launching a new rule. Your application must be ready for every scenario.

A custom application enables us to use specific modules we are interested in applying without adding complexity.

ExplorAI can help you to implement these systems in production. You can contact us for more details !

Explorai has been recognized as one of Top 30 Artificial Intelligence Agencies by DesignRush